Machine Learning and Economics

Tags:

Binary classification has been studied for a long time and is a classic machine-learning problem.

Many general-purpose metrics are available in that context that can be used to assess the performances of a predictive model: The Receiver Operating Characteristic (ROC) Curve and its associated Area Under the Curve (AUC) are among the most common.

However, the best model for the data scientist may not end up being the one maximizing the business value.

Nicolas Kruchten describes a practical case where the “best” model ends up being useless for the business owner. By adding the notion of expected utility into the mix, he shows how a less efficient model can lead to more profits.

This is an excellent article for any data scientist who wants to design models that add tangible business values.